How Otter.ai reached 35 million users by betting on voice AI early

Executive overview

Most voice knowledge throughout history has been lost — we never heard Shakespeare speak, never heard Darwin. Sam Liang built Otter.ai on the conviction that capturing and sharing spoken knowledge would become essential to how humans work.

Otter started as a transcription tool, evolved into an AI meeting assistant, and is now building a meeting-centric enterprise knowledge base. The hard bet: own the core speech technology rather than rely on third-party APIs.

Voice is becoming the primary interface for business intelligence — and 99% of the world hasn't adopted it yet.

Building on deep technology instead of APIs

• In 2016, anyone could build a basic meeting note-taker using existing APIs — that was the wrong path
• Owning speech recognition kept costs low, enabling more free-tier service than API-dependent competitors could offer
• Third-party APIs limit how much you can give away; cost structure becomes a ceiling on growth
• Unsolved problems remain: modelling multi-speaker conversations, using hundreds of millions of voice samples to understand human interaction — these require proprietary AI, not off-the-shelf tools
• If something is easy to build, a hundred others will build it too; the hard problems create defensible differentiation

Betting on behaviour change before it happens

• In 2016, being recorded felt uncomfortable; sharing meeting notes with colleagues was uncommon
• Otter built for the mindset shift they anticipated, not the mindset that existed
• Early adopters get value first, become more productive, then convert colleagues — adoption follows its own curve
• Today: people already use AI to write documents, emails, LinkedIn posts — the shift is underway and will accelerate
• In a few years, keyboards become secondary; talking to AI that writes for you becomes the norm

What enabled persistence over 10 years

• Stanford PhD advisor David Sheraton modelled long-range thinking about impact — the same advisor who backed Larry Page and Sergey Brin with $100k before Google existed
• Google Maps experience (2006–2010) gave Sam a foundation in large-scale location and behavioural systems
• His first startup (mobile location and persistent sensing) was acquired, validating the approach before Otter
• Running 11 marathons — and counting — as a deliberate strategy for managing stress and building through adversity
• Building a startup is harder than a marathon; most people quit early; the difficulties of ambitious projects are expected, not exceptional